Abstract
Elevated circulating cholesterol-rich low-density lipoprotein (LDL) particles increase coronary artery disease risk. Cell-surface hepatic LDL receptors (LDLRs) clear 70% of these particles from circulation. The ectodomain of LDLR is shed into circulation, preventing it from removing LDL particles. The role that LDLR ectodomain shedding plays as a regulatory mechanism is unknown.
We describe LDLR shedding via the relationships between circulating soluble LDLRs (sLDLRs) and serum lipoproteins, serum proprotein convertase subtilin/kexin type 9 (PCSK9; a negative regulator of LDLR), and clinical parameters in a white Canadian population.
Population-based, cross-sectional study.
Clinical Research Center, The Ottawa Hospital, and Faculty of Medicine, University of Ottawa.
Two hundred seventy-three white Canadians.
None.
sLDLR measured by ELISA; serum lipids and PCSK9, PCSK9 genotypes, and clinical parameters from previous analyses.
sLDLRs correlated strongly with triglycerides (TG; r = 0.624, P < 0.0001) and moderately with LDL cholesterol (r = 0.384, P < 0.0001), and high-density lipoprotein cholesterol (r = −0.307, P = 0.0003). Only TG correlations were unaffected by PCSK9 variations. sLDLR levels were significantly elevated in those with TG >50th or LDL cholesterol >75th percentiles.
Serum sLDLR levels correlate with several lipoprotein parameters, especially TG, and the presence of PCSK9 loss-of-function variants alters sLDLR levels and correlations, except for TG. Ectodomain LDLR shedding has a role in LDL metabolism, distinct from PCSK9, with interplay between these two pathways that regulate cell-surface LDLRs. Findings suggest alteration of LDLR shedding could emerge as a target to treat dyslipidemia.
Elevated low-density lipoprotein cholesterol (LDL-C) levels increase the risk of coronary artery disease, a leading cause of death worldwide (1). Approximately 70% of LDL particles are cleared by hepatic membrane-bound LDL receptors (LDLRs) (2). At the cell surface of the liver, LDLR binds the cholesterol-rich LDL particle through its protein component, apolipoprotein B100 (2). After endocytosis, the LDL particle is routed to the lysosome from the early-late endosome, and the LDLR is recycled to the cell surface, enabling multiple rounds of LDL particle binding, uptake, and clearance.
LDLR is a type 1 transmembrane N- and O-linked glycosylated protein. The amount of LDLR present on the hepatic cell surface is influenced by transcriptional and post-transcriptional factors. Transcriptionally, the cellular cholesterol status regulates LDLR gene expression through the sterol regulatory element binding protein-2, a transcription factor regulating genes involved in cholesterol biosynthesis, including the gene for the LDLR-regulating protein, PCSK9 (proprotein convertase subtilisin/kexin type 9) (3). When cellular cholesterol levels decrease, sterol regulatory element binding protein is activated, increasing LDLR expression and vice versa.
Post-transcriptionally, two well-recognized pathways that result in downregulation of cell-surface LDLRs include (1) downregulation via lysosomal degradation after LDLR C-terminal ubiquitination by the intracellular E3-ubiquitin ligase inducible degrader of LDLR (most active in extrahepatic tissues) (4); and (2) extracellular-mediated, PCSK9-directed LDLR downregulation (5). Most active at the hepatic cell surface, serum PCSK9 binds to an ectodomain of LDLR called the epidermal growth factor-like repeat A domain. Through an incompletely defined mechanism, the PCSK9–LDLR complex is shuttled to the lysosome for degradation (6). Recent research has identified liver cell surface proteoglycans as PCSK9 co-receptors, recruiting PCSK9 to the cell surface for LDLR interactions (7). More than 50 variants of PCSK9 have been identified with opposite effects on LDLR degradation: gain-of-function variants accelerate LDLR degradation and increase cholesterol levels and the risk of heart disease and loss-of-function (LOF) variants result in higher levels of LDLR, lower cholesterol levels, and protection from heart disease (8–10). LOF for several PCSK9 variants is attributed to a reduced ability to bind to cell-surface proteoglycans (7).
A third proposed mechanism to post-translationally regulate LDLR levels, and the least understood, is the shedding of the ectodomain of LDLR at the cell surface. This event involves cleavage of the LDLR near its transmembrane domain by undefined, extracellular matrix metalloproteases (11–13). This cleavage produces (1) an sLDLR molecule, composed of LDLR’s ectodomain; and (2) a membrane-bound stub, composed of its transmembrane and cytosolic domains, which is unable to uptake LDL particles. Shimohiro et al. (14) documented the association of sLDLR levels with lipoprotein parameters in 102 healthy Japanese adults. They found sLDLR strongly and positively associated with triglycerides (TG). Recently, Girona et al. (15) explored whether sLDLR could be used as a biomarker of familial hypercholesterolemia in children. It could not. However, they also reported a substantial correlation between the sLDLR and TG levels.
Despite knowledge of the phenomenon of LDLR shedding for more than one decade, very little is known about its effect on lipoprotein metabolism beyond the two cited studies, representing a substantial gap in our knowledge. Therefore, to understand the role of LDLR shedding in lipoprotein metabolism in an adult white Canadian population, we measured sLDLR levels in a cohort in which we have fully genotyped PCSK9, a known negative regulator of cell-surface LDLR, affording us the unique opportunity to document its effect on this process. In addition, we investigated whether the presence of PCSK9 LOF variations associated with lower LDL-C levels affect lipid association with sLDLR.
Subjects and Methods
Cohort study participants
A total of 273 participants gave informed written consent after approval of the study protocols by the Ottawa Hospital Research Institute and Clinical Research Institute of Montreal ethics committees. The study populations, described by Mayne et al. (16, 17), were not receiving any lipid-lowering medications. The lipid percentiles, which were age and sex matched, were determined using the Lipid Research Clinics table (18). Blood samples were obtained after a 12-hour fast. Lipids and lipoproteins were measured in accordance with the studies reported by Mayne et al. (16, 17). Genotyping of the exons of PCSK9 was performed in accordance with Mayne et al. (16). The various subgroup analyses performed based on sex, lipid percentiles, and PCSK9 genotype are shown in Supplemental Fig. 1. The PCSK9 variants identified in our white Canadian population are summarized in Supplemental Table 1. We excluded PCSK9 variants R46L, R93C, Q152H, E206K, V233L, and H553R from subgroup analyses (considered PCSK9 rare and other variants; n = 16) and multiple combinations of these variants (n = 5), because their occurrence was too low for statistical analyses.
sLDLR was measured in sera using a human LDLR ELISA from R&D Systems (Minneapolis, MN) with an intra-assay coefficient of variation (CV) of 2.0% to 2.2% and an interassay CV of 4.6% to 5.4%. The serum PCSK9 assay was performed using a human PCSK9 ELISA from CycLex Co. (Japan: MBL International, Woburn, MA) with an intra-assay CV of 1.5% to 2.6% and an interassay CV of 2.9% to 7.1%. All samples were quantified three times. The temporal stability of serum PCSK9 and LDLR levels were confirmed by ELISA using 10 individual serum samples undergoing repeated freeze-thaw cycles and measurement for 10 years for PCSK9 and 2 years for sLDLR.
Statistical analysis
Comparative data are presented as the median ± 10th to 90th percentile range. Results are expressed as the mean ± SEM, except as indicated. The data were tested for normality using the Kolmogorov-Smirnov test. The unpaired Student t test, Mann-Whitney U test or Kruskal-Wallis test (with the Dunn multiple comparison post test) was used for statistical analyses of differences, as appropriate and as indicated. A power analysis was performed to ensure comparisons were powered at >0.80 and α of 0.05 using PASS, version 12 (NCSS Statistical Software). Spearman correlation coefficients (r) were determined to assess the relationship between different parameters, because this nonparametric test allows for inclusion and analyses of skewed (PCSK9 and sLDLR) and outlier data. The data were analyzed using Prism, version 6, software (GraphPad, La Jolla, CA). Multivariable forward stepwise linear regression analyses were performed using SPSS software to determine the independent variables significantly associated with LDL-C and TG (IBM Corp., Armonk, NY). Correlation coefficients were compared using the Fisher r to z transformation using the following webpage: vassarstats.net/rdiff.html. Correlation coefficient values, r, were transformed into z scores before the comparison and analysis of statistical significance (z test statistic). Significance was defined as P < 0.05.
Results
Population characteristics
The population characteristics in terms of clinical lipoprotein and anthropometric measures are listed in Table 1, including PCSK9 and sLDLR measurements of 273 participants and subdivided by sex (men, n = 139; women, n = 134). Measures of PCSK9 and high-density lipoprotein cholesterol (HDL-C) were significantly higher in women than in men, and the body mass index (BMI) and TG levels were significantly higher in men than in women, as documented previously (16, 17). sLDLR levels did not differ significantly between the sexes.
Clinical Characteristics and Fasting Lipid, sLDLR, and PCSK9 Levels of Study Participants
| All (N = 273) | Men (n = 139) | Women (n = 134) | P Value | |
|---|---|---|---|---|
| sLDLR, ng/mL | 39.41 ± 1.53 (33.72, 1.91−185.0) | 41.39 ± 2.37 (39.47, 1.91−185.0) | 37.36 ± 1.92 (33.4, 2.04−179) | 0.541 |
| Age, y | 50.72 ± 0.89 (52, 18−91) | 51.04 ± 1.17 (53, 19−91) | 50.38 ± 1.37 (52, 18−89) | 0.777 |
| BMI, kg/m2 | 26.52 ± 0.30 (25.8, 15.8−45.1) | 27.51 ± 0.37 (26.8, 19.0−42.7) | 25.53 ± 0.44 (24.9, 15.8−45.1) | <0.0001a |
| PCSK9, ng/mL | 310.10 ± 7.89 (293.0, 51.2−858.7) | 285.2 ± 10.07 (254.1, 76.9−789.8) | 336.0 ± 11.85 (353.4, 51.2−858.7) | 0.0002a |
| TG, mM | 1.50 ± 0.044 (1.30, 0.37−4.08) | 1.80 ± 0.082 (1.54, 0.61−4.08) | 1.30 ± 0.046 (1.16, 0.55−3.64) | <0.0001a |
| TC, mM | 5.36 ± 0.069 (5.3, 2.8−8.7) | 5.31 ± 0.10 (5.2, 2.8−8.7) | 5.41 ± 0.092 (5.3, 3.37−8.30) | 0.389 |
| LDL-C, mM | 3.38 ± 0.064 (3.30, 1.10−6.10) | 3.40 ± 0.090 (3.39, 1.10−6.10) | 3.36 ± 0.089 (3.10, 1.13− 5.90) | 0.545 |
| HDL-C, mM | 1.30 ± 0.028 (1.22, 0.52−4.10) | 1.14 ± 0.031 (1.07, 0.52−2.53) | 1.47 ± 0.042 (1.32, 0.55−4.10) | <0.0001a |
| TC/HDL-C | 4.53 ± 0.099 (4.40, 1.80−11.10) | 5.03 ± 0.14 (5.00, 2.12−11.1) | 4.02 ± 0.12 (3.90, 1.80−10−0.20) | <0.0001a |
| Non–HDL-C, mM | 4.06 ± 0.071 (3.83, 1.68−7.92) | 4.17 ± 0.10 (4.03, 1.90−7.92) | 3.94 ± 0.097 (3.76, 1.68−6.73) | 0.106 |
| All (N = 273) | Men (n = 139) | Women (n = 134) | P Value | |
|---|---|---|---|---|
| sLDLR, ng/mL | 39.41 ± 1.53 (33.72, 1.91−185.0) | 41.39 ± 2.37 (39.47, 1.91−185.0) | 37.36 ± 1.92 (33.4, 2.04−179) | 0.541 |
| Age, y | 50.72 ± 0.89 (52, 18−91) | 51.04 ± 1.17 (53, 19−91) | 50.38 ± 1.37 (52, 18−89) | 0.777 |
| BMI, kg/m2 | 26.52 ± 0.30 (25.8, 15.8−45.1) | 27.51 ± 0.37 (26.8, 19.0−42.7) | 25.53 ± 0.44 (24.9, 15.8−45.1) | <0.0001a |
| PCSK9, ng/mL | 310.10 ± 7.89 (293.0, 51.2−858.7) | 285.2 ± 10.07 (254.1, 76.9−789.8) | 336.0 ± 11.85 (353.4, 51.2−858.7) | 0.0002a |
| TG, mM | 1.50 ± 0.044 (1.30, 0.37−4.08) | 1.80 ± 0.082 (1.54, 0.61−4.08) | 1.30 ± 0.046 (1.16, 0.55−3.64) | <0.0001a |
| TC, mM | 5.36 ± 0.069 (5.3, 2.8−8.7) | 5.31 ± 0.10 (5.2, 2.8−8.7) | 5.41 ± 0.092 (5.3, 3.37−8.30) | 0.389 |
| LDL-C, mM | 3.38 ± 0.064 (3.30, 1.10−6.10) | 3.40 ± 0.090 (3.39, 1.10−6.10) | 3.36 ± 0.089 (3.10, 1.13− 5.90) | 0.545 |
| HDL-C, mM | 1.30 ± 0.028 (1.22, 0.52−4.10) | 1.14 ± 0.031 (1.07, 0.52−2.53) | 1.47 ± 0.042 (1.32, 0.55−4.10) | <0.0001a |
| TC/HDL-C | 4.53 ± 0.099 (4.40, 1.80−11.10) | 5.03 ± 0.14 (5.00, 2.12−11.1) | 4.02 ± 0.12 (3.90, 1.80−10−0.20) | <0.0001a |
| Non–HDL-C, mM | 4.06 ± 0.071 (3.83, 1.68−7.92) | 4.17 ± 0.10 (4.03, 1.90−7.92) | 3.94 ± 0.097 (3.76, 1.68−6.73) | 0.106 |
Data presented as mean ± SEM (median, range).
Comparisons performed using the Mann-Whitney U test.
Statistically significant differences between men and women.
Clinical Characteristics and Fasting Lipid, sLDLR, and PCSK9 Levels of Study Participants
| All (N = 273) | Men (n = 139) | Women (n = 134) | P Value | |
|---|---|---|---|---|
| sLDLR, ng/mL | 39.41 ± 1.53 (33.72, 1.91−185.0) | 41.39 ± 2.37 (39.47, 1.91−185.0) | 37.36 ± 1.92 (33.4, 2.04−179) | 0.541 |
| Age, y | 50.72 ± 0.89 (52, 18−91) | 51.04 ± 1.17 (53, 19−91) | 50.38 ± 1.37 (52, 18−89) | 0.777 |
| BMI, kg/m2 | 26.52 ± 0.30 (25.8, 15.8−45.1) | 27.51 ± 0.37 (26.8, 19.0−42.7) | 25.53 ± 0.44 (24.9, 15.8−45.1) | <0.0001a |
| PCSK9, ng/mL | 310.10 ± 7.89 (293.0, 51.2−858.7) | 285.2 ± 10.07 (254.1, 76.9−789.8) | 336.0 ± 11.85 (353.4, 51.2−858.7) | 0.0002a |
| TG, mM | 1.50 ± 0.044 (1.30, 0.37−4.08) | 1.80 ± 0.082 (1.54, 0.61−4.08) | 1.30 ± 0.046 (1.16, 0.55−3.64) | <0.0001a |
| TC, mM | 5.36 ± 0.069 (5.3, 2.8−8.7) | 5.31 ± 0.10 (5.2, 2.8−8.7) | 5.41 ± 0.092 (5.3, 3.37−8.30) | 0.389 |
| LDL-C, mM | 3.38 ± 0.064 (3.30, 1.10−6.10) | 3.40 ± 0.090 (3.39, 1.10−6.10) | 3.36 ± 0.089 (3.10, 1.13− 5.90) | 0.545 |
| HDL-C, mM | 1.30 ± 0.028 (1.22, 0.52−4.10) | 1.14 ± 0.031 (1.07, 0.52−2.53) | 1.47 ± 0.042 (1.32, 0.55−4.10) | <0.0001a |
| TC/HDL-C | 4.53 ± 0.099 (4.40, 1.80−11.10) | 5.03 ± 0.14 (5.00, 2.12−11.1) | 4.02 ± 0.12 (3.90, 1.80−10−0.20) | <0.0001a |
| Non–HDL-C, mM | 4.06 ± 0.071 (3.83, 1.68−7.92) | 4.17 ± 0.10 (4.03, 1.90−7.92) | 3.94 ± 0.097 (3.76, 1.68−6.73) | 0.106 |
| All (N = 273) | Men (n = 139) | Women (n = 134) | P Value | |
|---|---|---|---|---|
| sLDLR, ng/mL | 39.41 ± 1.53 (33.72, 1.91−185.0) | 41.39 ± 2.37 (39.47, 1.91−185.0) | 37.36 ± 1.92 (33.4, 2.04−179) | 0.541 |
| Age, y | 50.72 ± 0.89 (52, 18−91) | 51.04 ± 1.17 (53, 19−91) | 50.38 ± 1.37 (52, 18−89) | 0.777 |
| BMI, kg/m2 | 26.52 ± 0.30 (25.8, 15.8−45.1) | 27.51 ± 0.37 (26.8, 19.0−42.7) | 25.53 ± 0.44 (24.9, 15.8−45.1) | <0.0001a |
| PCSK9, ng/mL | 310.10 ± 7.89 (293.0, 51.2−858.7) | 285.2 ± 10.07 (254.1, 76.9−789.8) | 336.0 ± 11.85 (353.4, 51.2−858.7) | 0.0002a |
| TG, mM | 1.50 ± 0.044 (1.30, 0.37−4.08) | 1.80 ± 0.082 (1.54, 0.61−4.08) | 1.30 ± 0.046 (1.16, 0.55−3.64) | <0.0001a |
| TC, mM | 5.36 ± 0.069 (5.3, 2.8−8.7) | 5.31 ± 0.10 (5.2, 2.8−8.7) | 5.41 ± 0.092 (5.3, 3.37−8.30) | 0.389 |
| LDL-C, mM | 3.38 ± 0.064 (3.30, 1.10−6.10) | 3.40 ± 0.090 (3.39, 1.10−6.10) | 3.36 ± 0.089 (3.10, 1.13− 5.90) | 0.545 |
| HDL-C, mM | 1.30 ± 0.028 (1.22, 0.52−4.10) | 1.14 ± 0.031 (1.07, 0.52−2.53) | 1.47 ± 0.042 (1.32, 0.55−4.10) | <0.0001a |
| TC/HDL-C | 4.53 ± 0.099 (4.40, 1.80−11.10) | 5.03 ± 0.14 (5.00, 2.12−11.1) | 4.02 ± 0.12 (3.90, 1.80−10−0.20) | <0.0001a |
| Non–HDL-C, mM | 4.06 ± 0.071 (3.83, 1.68−7.92) | 4.17 ± 0.10 (4.03, 1.90−7.92) | 3.94 ± 0.097 (3.76, 1.68−6.73) | 0.106 |
Data presented as mean ± SEM (median, range).
Comparisons performed using the Mann-Whitney U test.
Statistically significant differences between men and women.
Distribution of sLDLR levels in a white population
The distribution of sLDLR is right-skewed, with a median concentration of 33.72 ng/mL (mean ± SEM, 39.41 ± 1.53 ng/mL; Fig. 1). The median levels were not significantly different between the men (36.40 ng/mL) and women (33.40 ng/mL). sLDLR levels had a range of more than 150-fold, from a low of 1.91 ng/mL to a high of 185 ng/mL (Table 1).
Histogram of the distribution of sLDLR in a white Canadian population. Data are shown as relative frequency (percentage) for (A) all 273 participants, (B) 134 women, and (C) 139 men. The dotted vertical line indicates the median, and the solid vertical line, the mean values.
sLDLR is related to multiple lipoprotein and anthropometric characteristics in the Canadian white population
The Spearman correlations of sLDLR with age, BMI, PCSK9, and lipoprotein parameters are listed in Table 2. We found that sLDLR correlated positively with TG (r = 0.624; P < 0.0001), non–HDL-C (r = 0.4506; P < 0.0001), TC (r = 0.399; P < 0.0001), LDL-C (r = 0.384; P < 0.0001), BMI (r = 0.297; P < 0.0001), PCSK9 (r = 0.199; P = 0.0010), and age (r = 0.163; P = 0.0068) and correlated negatively with HDL-C (r = −0.307; P = 0.0003). For the men, all remained significantly correlated, except for age. For women, the correlation between sLDLR and PCSK9 was no longer statistically significant.
Associations Between sLDLR and Clinical Characteristics in Study Participants
| sLDLR (ng/mL) vs | All (N = 273) | Men (n = 139) | Women (n = 134) | |||
|---|---|---|---|---|---|---|
| r | P Value | r | P Value | r | P Value | |
| Age, y | 0.163 | 0.0068a | 0.0497 | 0.561 | 0.292 | 0.0006a |
| BMI, kg/m2 | 0.297 | <0.0001a | 0.244 | 0.0047a | 0.371 | <0.0001a |
| PCSK9, ng/mL | 0.199 | 0.0010a | 0.262 | 0.0019a | 0.163 | 0.061 |
| TG, mM | 0.624 | <0.0001a | 0.583 | <0.0001a | 0.718 | <0.0001a |
| TC, mM | 0.399 | <0.0001a | 0.369 | <0.0001a | 0.451 | <0.0001a |
| LDL-C, mM | 0.384 | <0.0001a | 0.315 | 0.0002a | 0.484 | <0.0001a |
| HDL-C, mM | −0.307 | 0.0003a | −0.253 | 0.0027a | −0.426 | <0.0001a |
| TC/HDL-C | 0.511 | <0.0001a | 0.469 | <0.0001a | 0.624 | <0.0001a |
| Non–HDL-C, mM | 0.506 | <0.0001a | 0.437 | <0.0001a | 0.601 | <0.0001a |
| sLDLR (ng/mL) vs | All (N = 273) | Men (n = 139) | Women (n = 134) | |||
|---|---|---|---|---|---|---|
| r | P Value | r | P Value | r | P Value | |
| Age, y | 0.163 | 0.0068a | 0.0497 | 0.561 | 0.292 | 0.0006a |
| BMI, kg/m2 | 0.297 | <0.0001a | 0.244 | 0.0047a | 0.371 | <0.0001a |
| PCSK9, ng/mL | 0.199 | 0.0010a | 0.262 | 0.0019a | 0.163 | 0.061 |
| TG, mM | 0.624 | <0.0001a | 0.583 | <0.0001a | 0.718 | <0.0001a |
| TC, mM | 0.399 | <0.0001a | 0.369 | <0.0001a | 0.451 | <0.0001a |
| LDL-C, mM | 0.384 | <0.0001a | 0.315 | 0.0002a | 0.484 | <0.0001a |
| HDL-C, mM | −0.307 | 0.0003a | −0.253 | 0.0027a | −0.426 | <0.0001a |
| TC/HDL-C | 0.511 | <0.0001a | 0.469 | <0.0001a | 0.624 | <0.0001a |
| Non–HDL-C, mM | 0.506 | <0.0001a | 0.437 | <0.0001a | 0.601 | <0.0001a |
Significance set at P < 0.05.
Statistically significant correlation.
Associations Between sLDLR and Clinical Characteristics in Study Participants
| sLDLR (ng/mL) vs | All (N = 273) | Men (n = 139) | Women (n = 134) | |||
|---|---|---|---|---|---|---|
| r | P Value | r | P Value | r | P Value | |
| Age, y | 0.163 | 0.0068a | 0.0497 | 0.561 | 0.292 | 0.0006a |
| BMI, kg/m2 | 0.297 | <0.0001a | 0.244 | 0.0047a | 0.371 | <0.0001a |
| PCSK9, ng/mL | 0.199 | 0.0010a | 0.262 | 0.0019a | 0.163 | 0.061 |
| TG, mM | 0.624 | <0.0001a | 0.583 | <0.0001a | 0.718 | <0.0001a |
| TC, mM | 0.399 | <0.0001a | 0.369 | <0.0001a | 0.451 | <0.0001a |
| LDL-C, mM | 0.384 | <0.0001a | 0.315 | 0.0002a | 0.484 | <0.0001a |
| HDL-C, mM | −0.307 | 0.0003a | −0.253 | 0.0027a | −0.426 | <0.0001a |
| TC/HDL-C | 0.511 | <0.0001a | 0.469 | <0.0001a | 0.624 | <0.0001a |
| Non–HDL-C, mM | 0.506 | <0.0001a | 0.437 | <0.0001a | 0.601 | <0.0001a |
| sLDLR (ng/mL) vs | All (N = 273) | Men (n = 139) | Women (n = 134) | |||
|---|---|---|---|---|---|---|
| r | P Value | r | P Value | r | P Value | |
| Age, y | 0.163 | 0.0068a | 0.0497 | 0.561 | 0.292 | 0.0006a |
| BMI, kg/m2 | 0.297 | <0.0001a | 0.244 | 0.0047a | 0.371 | <0.0001a |
| PCSK9, ng/mL | 0.199 | 0.0010a | 0.262 | 0.0019a | 0.163 | 0.061 |
| TG, mM | 0.624 | <0.0001a | 0.583 | <0.0001a | 0.718 | <0.0001a |
| TC, mM | 0.399 | <0.0001a | 0.369 | <0.0001a | 0.451 | <0.0001a |
| LDL-C, mM | 0.384 | <0.0001a | 0.315 | 0.0002a | 0.484 | <0.0001a |
| HDL-C, mM | −0.307 | 0.0003a | −0.253 | 0.0027a | −0.426 | <0.0001a |
| TC/HDL-C | 0.511 | <0.0001a | 0.469 | <0.0001a | 0.624 | <0.0001a |
| Non–HDL-C, mM | 0.506 | <0.0001a | 0.437 | <0.0001a | 0.601 | <0.0001a |
Significance set at P < 0.05.
Statistically significant correlation.
Additionally, to investigate the influence of age, we compared the sLDLR levels in men and women aged <50 and >50 years (Supplemental Fig. 2). No statistical differences were noted for men. sLDLR levels were significantly greater statistically in women aged >50 years compared with women aged <50 years (P = 0.0068). The 50-year age cutoff approximately divides premenopausal and perimenopausal women from postmenopausal women, suggesting that estrogen status might influence the process of LDLR shedding.
sLDLR levels increase with increasing TG and LDL-C percentiles
Box and whisker representations of the median and 10th to 90th percentile ranges are shown in Fig. 2A, with the mean indicated by the plus sign, for sLDLR as a function of TG subgroups [<5th (n = 9), 5th to 24th (n = 33), 25th to 49th (n = 69), 50th to 74th (n = 72), 75th to 95th (n = 75), and >95th (n = 15) percentiles]. No statistically significant difference for sLDLR levels less than the 49th percentile were found for TG. The 50th- to >95th-percentile TG subgroups had significantly higher sLDLR levels compared with all subgroups less than the 50th percentile. Additionally, the >95th-percentile LDL-C subgroup had significantly higher sLDLR levels compared with the 50th to 74th percentile.
Box and whisker representations of median ± 10th to 90th percentiles of sLDLR levels in a white Canadian population stratified by (A) TG and (B) LDL-C quartiles, with the extremes <5th and >95th shown. The mean is indicated by a plus sign. The TG and LDL-C percentiles were adjusted for age and sex. Comparisons were performed using the Kruskal-Wallis test with the Dunn multiple comparison post-test. Significance is indicated as *P < 0.05, ** P < 0.005, *** P < 0.0005, and **** P < 0.0001.
Figure 2B shows box and whisker representations of the median and 10th to 90th percentile ranges, with the mean indicated by the plus sign, for sLDLR as a function of LDL-C subgroups [< 5th (n = 30), 5th to 24th (n = 64), 25th to 49th (n = 66), 50th to 74th (n = 49), 75th to 95th (n = 45), and >95th (n = 19) percentiles]. Statistically significant differences among the subgroups were noted. No statistically significant differences were found for sLDLR levels less than the 74th percentile for LDL-C. The 75th- to 95th-percentile LDL-C subgroup had substantially higher sLDLR levels compared with all subgroups less than the 50th percentile, except for the <5th percentile. The >95th-percentile LDL-C subgroup had substantially higher sLDLR levels compared with all subgroups less than the 75th percentile.
Our results have shown that the strongest relationship between sLDLR and lipoprotein parameters is with the TG levels. We also tested whether the strength of this association differed between the upper and lower 50th percentiles for TGs (Supplemental Table 2). Similarly, we wanted to test whether the strength of this association differed between the upper and lower 50th percentiles of LDL-C, because (1) cell-surface LDLR plays a large role in the clearance of LDL from circulation, and (2) the relationship between cell surface and sLDLR is largely unknown (Supplemental Table 3). The positive correlation between sLDLR and TG, TC, and LDL-C did not differ between the <50th and >50th percentiles of TG using the Fisher r to z transformation to compare the correlation coefficients. However, division of the cohort between the <50th and >50th percentiles of LDL-C showed that although the strong association of sLDLR and TG remained, the sLDLR correlation with TC and LDL-C was substantially stronger in the >50th LDL-C percentile subgroup.
Additionally, we performed multivariable forward stepwise linear regression to determine the independent variables that were associated substantially with LDL-C and TG. In our cohort (n = 273), sLDLR [standardized β (sβ) = 0.294; cumulative R2 (cR2) = 0.140], PCSK9 (sβ = 0.263, cR2 = 0.201), age (sβ = 0.164, cR2 = 0.219), and HDL-C (sβ = −0.129, cR2 = 0.231) were independently associated with the LDL-C levels. These remained statistically significant when the analysis was performed on data from women (cR2 = 0.174, cR2 = 0.271, cR2 = 0.322, and cR2 = 0.373 for sLDLR, PCSK9, age, and HDL-C, respectively). In men, sLDLR (cR2 = 0.116) and PCSK9 (cR2 = 0.143) but not age or HDL-C were independently associated with the LDL-C levels.
In our cohort (n = 273), sLDLR (sβ = 0.494, cR2 = 0.328), HDL-C (sβ = −0.258, cR2 = 0.412), and sex (sβ = −0.173, cR2 = 0.423) were independently associated with the TG levels. Only sLDLR remained statistically significant when the analysis was performed on data from women (cR2 = 0.404). In the men, sLDLR (cR2 = 0.293) and HDL-C (cR2 = 0.437) were independently associated with the TG levels.
Effect of PCSK9 variation on sLDLR levels
We previously reported that PCSK9 LOF variants were associated with lower TC and LDL-C levels in a Canadian white cohort (16, 17). These included the PCSK9 I474V, L10A53V, R46L, and Q152H variants that, in that order, mildly to moderately to severely reduced the LDL-C levels (16, 17). The PCSK9 E670G variant was also studied; however, its classification remains controversial. Supplemental Table 1 shows the frequency of single (40.3%) and multiple (16.5%) PCSK9 exonic variations in the current white Canadian study population. Supplemental Table 4 shows the levels of TG, LDL-C, and TC in our carriers of single and multiple PCSK9 exonic variants compared with the nonvariant PCSK9 in our population. These analyses support our classification of L10A53V and I474V as LOF variants of PCSK9. However, the E670G did not result in substantially different TC and LDL-C levels compared with the non–PCSK9 variant population. Thus, we analyzed the PCSK9 E670G variant separately from our LOF PCSK9 variants. For subgrouping, other rare and uncommon PCSK9 single and multiple variants were excluded, because statistical analyses were precluded, as indicated in Supplemental Table 1.
Those carrying the PCSK9 variation (n = 155) had significantly lower sLDLR (median, 31.0 ng/mL) than those without PCSK9 variation (n = 118; median, 39.5 ng/mL; P = 0.0025; Fig. 3). Those carrying multiple LOF variations (n = 27; L10A53V and I474V) showed that this subgroup had significantly lower sLDLR levels (median, 26.8 ng/mL) from those without a PCSK9 variation (median, 39.5 ng/mL; P = 0.0022; Fig. 3). However, in those carrying the single LOF L10A53V (n = 42) or I474V (n = 47) PCSK9 variants, the sLDLR levels were not substantially lower than those without PCSK9 variants. The sLDLR levels in carriers of the E670G PCSK9 single variations were significantly lower than those in those without variants (27.0 ng/mL vs 39.4 ng/mL, respectively; P = 0.0066). This occurs without the substantial decrease in LDL-C noted for the LOF PCSK9 variants. Therefore, the presence of PCSK9 variation can be associated with a change in the level of circulating sLDLR, with or without an associated lowering of LDL-C.
Box and whisker representations of median ± 10th to 90th percentiles of sLDLR levels in a white Canadian population with and without PCSK9 exonic variations. The mean is indicated by a plus sign. Comparisons were performed using the Kruskal-Wallis test with the Dunn multiple comparison post-test. Significance is indicated as ** P < 0.005. NV, non-PCSK9 exon variant; LOF, carriers of either L10A53V or I474V and both L10A53V and I474V; Multiple LOF, carriers of any combination of L10A53V and I474V (Supplemental Table 1).
Effect of PCSK9 LOF on sLDLR’s relationship to multiple lipoprotein and anthropometric characteristics in the Canadian population
Because the presence of PCSK9 variations affected the circulating sLDLR levels (Fig. 3), we used Spearman correlations to explore the effect of the presence of PCSK9 variation on sLDLR’s relationship to the multiple lipoprotein and anthropometric characteristics in the Canadian population documented in Table 2 and Supplemental Table 1. Supplemental Tables 5 to 10 present Spearman correlations on subanalyses of our cohort according to the PCSK9 variant presence or absence, type of PCSK9 variation, whether it occurs singly or combined with another PCSK9 variation, and sex dichotomy.
Spearman correlation coefficients are summarized in Table 3 for these subgroups. The correlation between sLDLR and TG were the strongest, ranging from r = 0.470 to r = 0.728, whether in the absence or presence of PCSK9 variation. In addition, this association was not affected by sex. However, the remaining lipoprotein and clinical measure correlations were affected by the presence of PCSK9 variants. TC correlated with sLDLR for those not carrying PCSK9 variations; however, it was lost for those carrying PCSK9 LOF variants L10A53V and I474V. This association was not affected by sex. Both the positive correlation between sLDLR and LDL-C and its negative correlation with HDL-C were unaffected by sex in the non-PCSK9 variants. With PCSK9 variations, these associations differed. For those carrying PCSK9 LOF, only women carrying the L10A53V variant showed a correlation between sLDLR and LDL-C. In men, this association was lost. In women carrying LOF in PCSK9, the correlation between sLDLR and HDL-C remained, just as in the nonvariant population, but not for men carrying PCSK9 LOF; thus, a sex dichotomy was seen for those carrying PCSK9 LOF.
Summary Spearman Correlation Coefficients for Association of sLDLR and Lipoprotein, PCSK9 Levels, BMI, and Age in a Canadian White Population Stratified by Presence of PCSK9 Variant and Sex
| Group | Association of sLDLR With | ||||||
|---|---|---|---|---|---|---|---|
| TG | TC | LDL-C | HDL-C | PCSK9 | BMI | Age | |
| All (N = 273) | 0.624a | 0.399a | 0.384a | −0.307a | 0.199a | 0.297a | 0.163a |
| Men (n = 139) | 0.583a | 0.369a | 0.315a | −0.253a | 0.262a | 0.244a | 0.0497 |
| Women (n = 134) | 0.718a | 0.451a | 0.484a | −0.426a | 0.163 | 0.371a | 0.292a |
| Nonvariant (n = 118) | 0.633a | 0.416a | 0.388a | −0.284a | 0.103 | 0.239a | 0.0949 |
| Men (n = 72) | 0.647a | 0.426a | 0.406a | −0.306a | 0.124 | 0.244a | −0.00563 |
| Women (n = 46) | 0.665a | 0.394a | 0.371a | −0.359a | 0.0313 | 0.235 | 0.282 |
| L10A53V variant (n = 42) | 0.687a | 0.170 | 0.187 | −0.486a | 0.343a | 0.481a | 0.111 |
| Men (n = 17) | 0.635a | 0.191 | −0.034 | −0.331 | 0.436 | 0.419 | 0.043 |
| Women (n = 25) | 0.745a | 0.285 | 0.457a | −0.636a | 0.274 | 0.576a | 0.186 |
| I474V variant (n = 47) | 0.614a | 0.219 | 0.193 | −0.353a | 0.124 | 0.267 | 0.122 |
| Men (n = 25) | 0.588a | 0.177 | 0.0254 | −0.206 | 0.0285 | 0.155 | 0.233 |
| Women (n = 22) | 0.695a | 0.371 | 0.398 | −0.619a | 0.303 | 0.524a | 0.0955 |
| Multiple LOF variant (n = 27) | 0.470a | 0.211 | 0.241 | −0.101 | 0.116 | 0.349 | 0.472a |
| E670G variant (n = 16) | 0.565a | 0.525a | 0.536a | −0.160 | 0.321 | −0.209 | −0.532 |
| Group | Association of sLDLR With | ||||||
|---|---|---|---|---|---|---|---|
| TG | TC | LDL-C | HDL-C | PCSK9 | BMI | Age | |
| All (N = 273) | 0.624a | 0.399a | 0.384a | −0.307a | 0.199a | 0.297a | 0.163a |
| Men (n = 139) | 0.583a | 0.369a | 0.315a | −0.253a | 0.262a | 0.244a | 0.0497 |
| Women (n = 134) | 0.718a | 0.451a | 0.484a | −0.426a | 0.163 | 0.371a | 0.292a |
| Nonvariant (n = 118) | 0.633a | 0.416a | 0.388a | −0.284a | 0.103 | 0.239a | 0.0949 |
| Men (n = 72) | 0.647a | 0.426a | 0.406a | −0.306a | 0.124 | 0.244a | −0.00563 |
| Women (n = 46) | 0.665a | 0.394a | 0.371a | −0.359a | 0.0313 | 0.235 | 0.282 |
| L10A53V variant (n = 42) | 0.687a | 0.170 | 0.187 | −0.486a | 0.343a | 0.481a | 0.111 |
| Men (n = 17) | 0.635a | 0.191 | −0.034 | −0.331 | 0.436 | 0.419 | 0.043 |
| Women (n = 25) | 0.745a | 0.285 | 0.457a | −0.636a | 0.274 | 0.576a | 0.186 |
| I474V variant (n = 47) | 0.614a | 0.219 | 0.193 | −0.353a | 0.124 | 0.267 | 0.122 |
| Men (n = 25) | 0.588a | 0.177 | 0.0254 | −0.206 | 0.0285 | 0.155 | 0.233 |
| Women (n = 22) | 0.695a | 0.371 | 0.398 | −0.619a | 0.303 | 0.524a | 0.0955 |
| Multiple LOF variant (n = 27) | 0.470a | 0.211 | 0.241 | −0.101 | 0.116 | 0.349 | 0.472a |
| E670G variant (n = 16) | 0.565a | 0.525a | 0.536a | −0.160 | 0.321 | −0.209 | −0.532 |
Supplemental Tables 5 through 10 show individual associations, with P values reported.
Statistically significant correlation between the variable and sLDLR (P < 0.05).
Summary Spearman Correlation Coefficients for Association of sLDLR and Lipoprotein, PCSK9 Levels, BMI, and Age in a Canadian White Population Stratified by Presence of PCSK9 Variant and Sex
| Group | Association of sLDLR With | ||||||
|---|---|---|---|---|---|---|---|
| TG | TC | LDL-C | HDL-C | PCSK9 | BMI | Age | |
| All (N = 273) | 0.624a | 0.399a | 0.384a | −0.307a | 0.199a | 0.297a | 0.163a |
| Men (n = 139) | 0.583a | 0.369a | 0.315a | −0.253a | 0.262a | 0.244a | 0.0497 |
| Women (n = 134) | 0.718a | 0.451a | 0.484a | −0.426a | 0.163 | 0.371a | 0.292a |
| Nonvariant (n = 118) | 0.633a | 0.416a | 0.388a | −0.284a | 0.103 | 0.239a | 0.0949 |
| Men (n = 72) | 0.647a | 0.426a | 0.406a | −0.306a | 0.124 | 0.244a | −0.00563 |
| Women (n = 46) | 0.665a | 0.394a | 0.371a | −0.359a | 0.0313 | 0.235 | 0.282 |
| L10A53V variant (n = 42) | 0.687a | 0.170 | 0.187 | −0.486a | 0.343a | 0.481a | 0.111 |
| Men (n = 17) | 0.635a | 0.191 | −0.034 | −0.331 | 0.436 | 0.419 | 0.043 |
| Women (n = 25) | 0.745a | 0.285 | 0.457a | −0.636a | 0.274 | 0.576a | 0.186 |
| I474V variant (n = 47) | 0.614a | 0.219 | 0.193 | −0.353a | 0.124 | 0.267 | 0.122 |
| Men (n = 25) | 0.588a | 0.177 | 0.0254 | −0.206 | 0.0285 | 0.155 | 0.233 |
| Women (n = 22) | 0.695a | 0.371 | 0.398 | −0.619a | 0.303 | 0.524a | 0.0955 |
| Multiple LOF variant (n = 27) | 0.470a | 0.211 | 0.241 | −0.101 | 0.116 | 0.349 | 0.472a |
| E670G variant (n = 16) | 0.565a | 0.525a | 0.536a | −0.160 | 0.321 | −0.209 | −0.532 |
| Group | Association of sLDLR With | ||||||
|---|---|---|---|---|---|---|---|
| TG | TC | LDL-C | HDL-C | PCSK9 | BMI | Age | |
| All (N = 273) | 0.624a | 0.399a | 0.384a | −0.307a | 0.199a | 0.297a | 0.163a |
| Men (n = 139) | 0.583a | 0.369a | 0.315a | −0.253a | 0.262a | 0.244a | 0.0497 |
| Women (n = 134) | 0.718a | 0.451a | 0.484a | −0.426a | 0.163 | 0.371a | 0.292a |
| Nonvariant (n = 118) | 0.633a | 0.416a | 0.388a | −0.284a | 0.103 | 0.239a | 0.0949 |
| Men (n = 72) | 0.647a | 0.426a | 0.406a | −0.306a | 0.124 | 0.244a | −0.00563 |
| Women (n = 46) | 0.665a | 0.394a | 0.371a | −0.359a | 0.0313 | 0.235 | 0.282 |
| L10A53V variant (n = 42) | 0.687a | 0.170 | 0.187 | −0.486a | 0.343a | 0.481a | 0.111 |
| Men (n = 17) | 0.635a | 0.191 | −0.034 | −0.331 | 0.436 | 0.419 | 0.043 |
| Women (n = 25) | 0.745a | 0.285 | 0.457a | −0.636a | 0.274 | 0.576a | 0.186 |
| I474V variant (n = 47) | 0.614a | 0.219 | 0.193 | −0.353a | 0.124 | 0.267 | 0.122 |
| Men (n = 25) | 0.588a | 0.177 | 0.0254 | −0.206 | 0.0285 | 0.155 | 0.233 |
| Women (n = 22) | 0.695a | 0.371 | 0.398 | −0.619a | 0.303 | 0.524a | 0.0955 |
| Multiple LOF variant (n = 27) | 0.470a | 0.211 | 0.241 | −0.101 | 0.116 | 0.349 | 0.472a |
| E670G variant (n = 16) | 0.565a | 0.525a | 0.536a | −0.160 | 0.321 | −0.209 | −0.532 |
Supplemental Tables 5 through 10 show individual associations, with P values reported.
Statistically significant correlation between the variable and sLDLR (P < 0.05).
Correlations for those carrying the E670G PCSK9 variant mirrored that of the non-PCSK9 variant population, except for the sLDLR association with HDL-C, which was lost in the population carrying the PCSK9 E670G variation. The correlation of PCSK9 with sLDLR resulted from the presence of the L10A53V PCSK9 variant. sLDLR levels correlated with the BMI in the non–PCSK9 variant men but did so with the women carrying PCSK9 LOF. Finally, sLDLR did not correlate with age in those carrying single PCSK9 variants but was seen in women without the PCSK9 variation. Therefore, the presence of PCSK9 LOF variants negated sLDLR’s association with TC and LDL-C (except in women carrying L10A53V) and the presence of the E670G PCSK9 variant negated the association between sLDLR and HDL-C (Table 3).
Discussion
The two major findings from the present study are that (1) serum sLDLR levels correlate with several serum lipoprotein parameters, especially TG, suggesting that ectodomain shedding of the LDLR affects lipoprotein metabolism; and (2) the presence of PCSK9 LOF variants alters sLDLR levels and sLDLR’s correlations with lipoprotein parameters, suggesting an interplay between two post-transcriptional regulators of LDLR function: PCSK9 and LDLR shedding.
The relationship and the significance of the relationship between sLDLR and lipoprotein metabolism are largely unknown (summarized in Supplemental Fig. 3). Because cell membrane-anchored LDLR plays a key role in the regulation of LDL clearance from circulation, it is plausible that its ectodomain shedding would have an effect on LDL clearance. Our data showing a correlation of sLDLR with TC and LDL-C are consistent with this possibility. When our cohort was stratified according to LDL-C percentiles, the higher percentile groups had higher serum sLDLR levels and stronger correlations of sLDLR with TC and LDL-C (Supplemental Table 3). These findings suggest that LDLR shedding leads to diminished LDL clearance. The mechanisms involved are not clear. It might be that LDLR shedding results in a decrease in LDL clearance through a reduction in the number of cell-surface LDLRs (Supplemental Fig. 3, 1). It is also possible that sLDLR acts as a competitive binder of LDL particles in circulation, resulting in decreased LDL clearance via cell-surface LDLR (Supplemental Fig. 3, 2). The correlations between sLDLR and various lipoprotein parameters according to our whole cohort are summarized in Table 2. Our results have both similarities and differences to those of Shimohiro et al. (14) in a Japanese cohort. Both studies showed moderate correlation between sLDLR and LDL-C. However, when men and women were analyzed separately, the correlation remained statistically significant for women in both cohorts but for men only in our all-white cohort.
The positive correlation of sLDLR with TG is especially remarkable in that it is less expected and yet much stronger than with LDL-C. It is now the third time this has been demonstrated. The first was in the study by Shimohiro et al. (14) in the Japanese cohort (r = 0.408) and the second in 177 children with and without familial hypercholesterolemia reported by Girona et al. (15) (r = 0.298). In our larger white cohort, the correlation with TG was stronger (r = 0.624) by z stat compared with the previous two studies. This consistent finding across studies raises the possibility of an effect of LDLR shedding on the metabolism of remnant lipoproteins. The reduction in cell-surface LDLR through shedding could be the reason for the diminished clearance of lipoprotein remnants. However, remnant lipoproteins are mainly cleared by the LDLR-related protein-1 (LRP) pathway (19). The LDLR deficiency seen in familial hypercholesterolemia is usually not associated with hypertriglyceridemia. Thus, the positive correlation between sLDLR and TG might be an epiphenomenon resulting from a concomitant shedding of LRP causing TG changes. Similar to LDLR, LRP is known to undergo ectodomain shedding (19). Another alternative suggested by Shimohiro et al. (14) was that of hepatic fat accumulation in hypertriglyceridemic subjects, resulting in induction of TNF-α and activation of a matrix metalloproteinase such as ADAM17 at the cell surface for cleavage of LDLR (Supplemental Fig. 3, 3).
In addition to correlations with LDL-C and TG, sLDLR also correlated (negatively) with HDL-C in our Canadian cohort. This was not seen in the studies by Shimohiro et al. (14) and Girona et al. (15). Overall, the important correlations of sLDLR with multiple lipoprotein parameters point to the possibility of LDLR shedding having an effect on lipoprotein metabolism through more than one mechanism (Supplemental Fig. 3). The underlying mechanisms that lead to these relationships and differences between populations are unknown but deserve further exploration. The differences noted between our study and the two cited studies might have resulted, in part, to differences in the frequency of PCSK9 gene variations in our respective populations, as reported by Mayne et al. (17) and Miyake et al. (20), and the effect of those PCSK9 variants as a LOF and gain-of-function on LDLR/LDL-C metabolism.
To the best of our knowledge, no study has reported on the effect of PCSK9 variations on sLDLR levels and the correlation with metabolic factors. Our cohort is well characterized in terms of PCSK9 in that all exons of PCSK9 were sequenced and analyzed for variation (16, 17). Most PCSK9 variants in our subjects were LOF (including the frequent L10A53V and I474V and infrequent R46L, Q152H, and R93C) or neutral (E670G), with a high frequency of multiplicity of variants (16.5%; Supplemental Table 1). We found that although those with a single L10A53V or I474V variant did not have any difference in sLDLR levels from those without variants, those with more than one LOF variant in PCSK9 had significantly lower levels of circulating sLDLR than those without variants (Fig. 3). Also, when all subjects with a single variant or multiple variants were grouped, they had lower sLDLR levels compared with the nonvariant group. These data suggest that PCSK9 LOF has an influence on circulating sLDLR levels. Kwakernaak et al. (21) found that circulating PCSK9 levels were strongly correlated with the TG-rich intermediate density lipoprotein (IDL) subfraction of LDL and not the small and large subfractions of LDL, suggesting that PCSK9 might affect TG via its effect on the metabolism of TG-rich LDL subfractions (such as IDL) (21). Future analyses should include subfractionation studies.
The strong correlation between sLDLR and TG levels shown in our subgroup analyses was unaffected by the presence of PCSK9 LOF variants, regardless of sex (Table 3). In contrast, the correlation between sLDLR and LDL-C was negated by the presence of PCSK9 LOF variants, with only one exception, women with L10A53V. The negative correlation between sLDLR and HDL-C was retained in women but lost in men carrying either the L10A53V or I474V PCSK9 variation. These findings suggest that the mechanisms involved in the interplay between LDLR shedding and PCSK9 LOF are different for TG, LDL-C, and HDL-C. In contrast to LOF PCSK9 variants, the neutral PCSK9 variant, E670G, showed no effect on the correlation between sLDLR and TG, TC, and LDL-C, just as in the nonvariant population, but negated the correlation with HDL-C. No sex dichotomy was found in the PCSK9 nonvariants with regard to sLDLR levels and its association with TG, TC, LDL-C, and HDL-C. However, in the presence of an LOF variant (L10A53V or I474V), a clear sex dichotomy was found in the correlation between sLDLR and HDL-C. Sex hormones might play a role in sLDLR formation. Estrogen affects LDLR expression at the transcriptional level, and we have also shown that estrogen affects post-translational regulation of LDLR through PCSK9 (22, 23). This should be further studied.
Nonvariants did not show an association between sLDLR and PCSK9 levels, such as was observed for the whole cohort. It is unknown whether PCSK9, and its variants, directly bind to sLDLR in circulation, or how this would affect the function of circulating sLDLR and the parameters we have discussed (Supplemental Fig. 3, 4). The correlation of plasma PCSK9 with sLDLR was attributed to the presence of the LOF PCSK9 L10A53V variant. Those carrying the L10A53V PCSK9 variant have PCSK9 levels comparable or slightly higher than nonvariants despite its LOF (17). The L10A53V PCSK9 variant was also associated with reduced circulating levels of the phosphorylated form of PCSK9 compared with nonvariants (22). A combination of these effects might be reflected in the gain of correlation between PCSK9 and sLDLR in these carriers.
In the present study, we noted a strong correlation between sLDLR and TG and a moderate association with LDL-C. This suggests that sLDLR might primarily affect TG-rich lipoprotein particles, with a secondary effect on LDL particles, and PCSK9 primarily affects LDL-C particle levels through LDLR downregulation, with secondary effects on TG metabolism. An important limitation particular to our cross-sectional study was that although interesting associations and correlations have been established, the cause and effect relationships could not be determined. Additionally, experimental studies such as pulse-chase analyses of soluble LDLR–LDLR conversion and apolipoprotein B100 clearance studies in relation to soluble LDLR are required to understand the role of sLDLR in LDL and TG metabolic pathways. Nonetheless, the present study has provided useful information in an area that has many gaps in knowledge. Our clinical observations provide useful clues to guide further investigation, which can lead to the discovery of methods to regulate LDLR shedding as a novel dyslipidemic target.
Conclusion
Serum sLDLR levels correlate with several lipoprotein parameters, especially TG, and the presence of PCSK9 LOF variants alters sLDLR levels and correlations, except for TG. Ectodomain shedding of LDLR has a role in LDL metabolism, distinct from that of PCSK9, with interplay between these two pathways that regulate cell-surface LDLR. These findings suggest that alteration of LDLR shedding can emerge as a target to treat dyslipidemia.
Abbreviations:
- BMI
body mass index
- cR2
cumulative R2
- CV
coefficient of variation
- HDL-C
high-density lipoprotein cholesterol
- LDL
low-density lipoprotein
- LDL-C
low-density lipoprotein cholesterol
- LDLR
low-density lipoprotein receptor
- LOF
loss-of-function
- LRP
low-density lipoprotein receptor-related protein-1
- PCSK9
proprotein convertase subtilisin/kexin type 9
- sLDLR
soluble low-density lipoprotein receptor
- sβ
standardized β
- TC
total cholesterol
- TG
triglycerides
Acknowledgments
The authors thank their participants.
Financial Support: This work was supported by the Heart and Stroke of Canada (grant G-16-00014693 to D.F. and J.M. and grant NA-7278 to T.C.O.) and the La Foundation J-Louis Lévesque (to D.F. and M.C.).
AuthorContributions: J.M., T.C.O., H.A., and D.F. conceived the study, directed the research, and wrote the manuscript. T.C.O., M.M., H.W., H.A., and M.C. recruited the participants. J.M., L.T., K.W., Z.N., D.S., and D.M. performed our measurement and data analyses. All the authors participated in figure and table preparation, edited the manuscript, and approved the final version.
Disclosure Summary: The authors have nothing to disclose.
